Vowels in chains

I don’t like vowels much. They are slippery little beasts. They wander about when you’re not looking. Give me a good, honest consonant any day.

There are some well-documented shifts in vowel quality going on now, and some that have been in progress for quite a time. One is the Northern Cities Shift in the USA.

Another is front vowel raising in NZ English, whereby the TRAP vowel is raised to about the quality of the GBE DRESS vowel, NZ DRESS sounds a lot like GBE KIT, and NZ KIT sounds like a raised schwa. The raising is even more extreme for some speakers. Almost 20 years ago in Wellington I was somewhat surprised at a shop assistant’s utterance as she handed me a $10 note in my change. She said: “Here’s your tiːn“. The well-known NZ SQUARE-NEAR merger used to result in a quality something like , but these days the quality is usually ɪə.

I think there is a chain shift going on in GBE too, but in the opposite direction to the NZ changes. The change in quality of the GOOSE vowel from u to ʉ, sometimes accompanied by unrounding, has been commented on for quite a few years now. I have recently been noticing younger speakers producing a vowel in <ing> endings which sounds very like the DRESS vowel, so for example shopping sounds a lot like ˈʃɒpeŋ. Finally, the DRESS vowel is often lowered to sound rather like æ, especially after w. I have lost count of the times weather forecasters have warned me of impending ˈwæt ˈwæðə. Put all these together and you get something like what is diagrammed below.

8 thoughts on “Vowels in chains

  1. You’re talking about chains, and the THOUGHT vowel and the LOT vowel, which I often hear as [o] (French eau), are certainly pushing GOOSE and FOOT, or was sucked there by the nearby underpressure.

  2. Phillip,

    Yes you’re right. And I deliberately avoided committing myself on whether the chains I mentioned are push or pull types. It’s quite possible of course for a shift to be of both types simultaneously.

  3. Regarding brighter pronunciations of the GOOSE vowel, phonetic descriptions of French, German and Swedish have always emphasized the darkness of their respective /u/ compared to English, and English speaking learners have always been advised to darken /u/ completely in those languages (as far back as the 19th c at least). So this is is hardly new. I’ve probably been doing it all my life, documented in the 1960s. My F2 ranged from 1200 to 1600Hz in /u/ in continuous speech, while a politician doing a political speech had 900 to 1200Hz (both NE Kent). Published data from French, German and Swedish usually show around 800-1000Hz or lower. Part of the difference is actual degree of lip rounding (less in English, necessarily more for French. German and Swedish). The rest isn’t tongue body fronting, the velar constriction is maintained. In those two English examples, the darkest /u/ allophones were after labial consonants, fairly dark after velars, brighter after dentals. X-ray profiles from all these languages show the tongue blade curled down tightly in French, German and Swedish, but not in English. So the other factor in brighter instances of English /u/ is tongue blade elevation.

  4. Sidney,

    According to Wikipedia, Swedish has a compressed close back vowel rather than a protruded one. Is this true in your experience? Or does it actually vary? Also do you think that lip protrusion would lower F2 more than lip compression?

  5. Pat, sorry about the delay, I wanted to check Wikipedia and do some homework on the spectral effects of lip activity. I found http://en.wikipedia.org/wiki/Swedish_phonology, which could be the page you read.

    Swedish vowels first. Swedish, like Norwegian, manages to pack in four contrasting vowel phonemes with low F1: /i y ʉ u/, sharing different portions of the F2 range (from high to low). Amazingly, /ʉ/ has a similar palatal tongue position as /y ø/. Finland Swedish doesn’t distinguish /ʉ/ from /u/ (hope that’s not oversimplified). All can be long or short. Regional accents add in their own local mix of laxing for the short vowels (less lip activity, modified tongue posture etc) raising F1 somewhat (but /ø/ isn’t far away). The long vowels tend to be diphthongs, lingual for /i: y:/ [iʝ yɥ], labial for /ʉ: u:/ [ʉβ uβ], adding redundancy to this slender contrast. This is the first part of what Wikipedia refers to as compression in /u/: the tendency to progressively narrow the lip opening towards [β].

    This means there are two vowels, /y ʉ/, that differ primarily in lip activity (traditionally known as out-rounding and in-rounding in Swedish phonetics), to keep their respective F2 ranges separate. Now we need to distinguish between lip approximation (lips moving directly towards each other, e.g. for [p]) and lip protrusion (lips pouting outwards as for a kiss). Swedish /u/ needs both protrusion and approximation, simultaneously lengthening and narrowing the opening. But Swedish /y/ gets only protrusion, lengthening but not narrowing the opening, so F2 drops a little (away from /i/) but not as far as for /ʉ/. Finally, /ʉ/ gets only approximation, narrowing the lip opening but not lengthening it, lowering F2 away from /y/ but still higher than for /u/. So compression for /ʉ/ means (i) no lip protrusion, keeping it away from /y/, and (ii) extreme approximation, leading ultimately to turbulent [β], like for /u/.

    That’s also a first answer to your second question, about the respective effects of lip protrusion and compression on F2. But it’s not that simple, Swedish /y/ out-rounding is relatively open, Swedish /ʉ/ in-rounding is extremely narrow.

  6. Pat, the second question: ‘Also do you think that lip protrusion would lower F2 more than lip compression?’

    Any effect must depend on the magnitude of the one or the other. And possibly also varying depending on the internal configuration of the vocal tract. Lip approximation can be made so small that turbulence is created, like [β]. Actual physical size would limit protrusion. One person might manage a mm or two more then the next one, and so lower F2 a little more.

    Stevens and House, and Fant, modelled the lip opening in the 1950s as A cm²/l cm, representing the acoustic impedance of the mouth opening, where A is the cross section area of the lip opening, and l its length. Assuming a fixed mandible, A will vary with lip approximation alone. Roughly speaking, values of A/l about 2-6cm correspond to spread lips, and smaller than 1cm correspond to rounded lips. Playing with this model might help.

    Suppose A is 0.12cm² (a lip opening around 4mm diameter) and l is 1cm, for an [u]-like vowel. A/l=0.12cm. Changing A to 0.2cm² (lip opening around 5mm diameter, i.e. 1mm less approximation), the lip protrusion has to be increased to 1.6cm to keep the same A/l value 0.12cm and the same acoustic effect. In that example, 1mm of approximation appears to be worth 6mm of protrusion. If your interested in articulatory compensation, that’s what it could be about. Like Jespersen’s pipe smoker. If the pipe holds the lip approximation back by 1mm, he needs an extra 6mm protrusion for that [u]. Not quite what you asked, but a step down the road.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.